20230511_ハードコーディングを避けたスクリプト.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# データの読み込み"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-05-11T09:36:24.660889Z",
     "start_time": "2023-05-11T09:36:24.215634Z"
    }
   },
   "outputs": [],
   "source": [
    "# トイデータを読み込む\n",
    "import pandas as pd\n",
    "df = pd.read_csv(\"toy_data.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AdaBoostの実装"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ハードコーディングして解消したコード"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-05-11T09:36:24.685457Z",
     "start_time": "2023-05-11T09:36:24.660889Z"
    }
   },
   "outputs": [],
   "source": [
    "import math\n",
    "import numpy as np\n",
    "\n",
    "class AdaboostHandmade:\n",
    "    '''\n",
    "    n_estimetors:学習する弱分類器の個数\n",
    "    weak_learner:弱分類器（DecisionTreeClassifier）を想定\n",
    "    '''\n",
    "    \n",
    "    def __init__(self, n_estimators, WeakLearner, **params):\n",
    "        self._n_estimators = n_estimators\n",
    "        self._WeakLearner = WeakLearner\n",
    "        self._params = params\n",
    "        \n",
    "    \n",
    "    @property\n",
    "    def n_estimators(self):\n",
    "        return self._n_estimators\n",
    "    \n",
    "    @property\n",
    "    def WeakLearner(self):\n",
    "        return self._WeakLearner\n",
    "    \n",
    "    @property\n",
    "    def params(self):\n",
    "        return self._params\n",
    "    \n",
    "    def fit(self, X, y):\n",
    "        # 弱分類器を保存するためのリスト\n",
    "        self.weak_learner_list = []\n",
    "        # alpha_mを保存するためのリスト\n",
    "        self.alpha_m_list = []\n",
    "        \n",
    "        # 重み\n",
    "        weight = [1/len(X)]*len(X)\n",
    "        weight = np.array(weight)\n",
    "        \n",
    "        # 弱分類器を連続して作成する\n",
    "        for m in range(self.n_estimators):\n",
    "            # 弱分類器のインスタンスを作成する\n",
    "            weak_learner = self._WeakLearner()\n",
    "            weak_learner.set_params(**self._params)\n",
    "            \n",
    "            # 弱分類器の学習\n",
    "            weak_learner.fit(X, y, sample_weight=weight)\n",
    "            self.weak_learner_list.append(weak_learner)\n",
    "            # err_mを計算する\n",
    "            y_pred = weak_learner.predict(X)\n",
    "            err_rate = (y != y_pred).astype(\"float\")\n",
    "            err_m_numerator =  np.dot(weight, err_rate)\n",
    "            err_m_denominator = np.sum(weight)\n",
    "            err_m = err_m_numerator/err_m_denominator\n",
    "            \n",
    "            # alpha_mを計算する\n",
    "            self.alpha_m = math.log((1-err_m)/err_m)\n",
    "            self.alpha_m_list.append(self.alpha_m)\n",
    "            \n",
    "            # 重みを更新する\n",
    "            weight = weight*np.array([math.exp(tmp) for tmp in self.alpha_m*err_rate])\n",
    "            ## 重みの正則化\n",
    "            weight = weight/np.sum(weight)\n",
    "            \n",
    "        return self.weak_learner_list\n",
    "        \n",
    "    def predict(self, X):\n",
    "        pred_y_rate_list = np.array([weak_learner.predict(X) for weak_learner in self.weak_learner_list])\n",
    "        \n",
    "        # 弱分類器が異なることをわかりやすくするためのprint文\n",
    "        for weak_learner in self.weak_learner_list:\n",
    "            print(\"predict:\", weak_learner.predict(X))\n",
    "        \n",
    "        pred_y_rate_list_decode = 2*pred_y_rate_list-1\n",
    "        pred_y_rate_list_encode = np.apply_along_axis(lambda gx:np.dot(self.alpha_m_list, gx) , axis=0, arr=pred_y_rate_list_decode) # 自由度カイ二乗分布にしたがう乱数の作成        \n",
    "        pred_y_list_decode = (pred_y_rate_list_encode>=0).astype(\"float\")\n",
    "        \n",
    "        return pred_y_list_decode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-05-11T09:36:26.644583Z",
     "start_time": "2023-05-11T09:36:24.689475Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 1. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 1. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 1. 0. ... 0. 1. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 1. ... 1. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 1. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 1. ... 0. 1. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [1. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [1. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 1. 0. ... 0. 1. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 1. ... 0. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 1. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [0. 0. 0. ... 0. 0. 1.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 1. 1. ... 1. 1. 1.]\n",
      "predict: [0. 0. 0. ... 0. 0. 0.]\n",
      "predict: [1. 0. 0. ... 0. 0. 1.]\n",
      "acurracy_score: 0.843\n",
      "confusion_matrix:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[966,  49],\n",
       "       [265, 720]], dtype=int64)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "features_name = [\"X1\", \"X2\", \"X3\", \"X4\", \"X5\", \"X6\", \"X7\", \"X8\", \"X9\", \"X10\"]\n",
    "adaboost_handmade = AdaboostHandmade(n_estimators=100, WeakLearner=DecisionTreeClassifier, max_depth=1, random_state=501)\n",
    "weak_lerner_list = adaboost_handmade.fit(X=df[features_name], y=df[\"label\"])\n",
    "pred_labels = adaboost_handmade.predict(X=df[features_name])\n",
    "\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix\n",
    "print(\"acurracy_score:\", accuracy_score(df[\"label\"], pred_labels))\n",
    "print(\"confusion_matrix:\")\n",
    "display(confusion_matrix(df[\"label\"], pred_labels))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}